Imports

from plotnine import *

from sklearn.preprocessing import StandardScaler

from sklearn.cluster import KMeans
from sklearn.mixture import GaussianMixture

from sklearn.metrics import silhouette_score

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import warnings
import plotly.express as px
import seaborn as sns
import plotly.graph_objects as go
from plotly.subplots import make_subplots
import math
from numpy.linalg import norm
from sklearn.cluster import DBSCAN
from sklearn.neighbors import NearestNeighbors
%matplotlib inline

sns.set()
warnings.filterwarnings("ignore")

Functions

def multiple_histogram(data):
    
    num_row = math.ceil(len(data.columns) / 3)
    fig = make_subplots(rows=num_row, cols=3,subplot_titles=(df.columns))

    place_col = 1
    place_row = 1
    for i in data.columns:

        fig.add_trace(go.Histogram(x=data[i]),row=place_row, col=place_col)

        place_col += 1
        if place_col == 4:
            place_col = 1
            place_row += 1
    return fig.update_layout(height=1600, width=1100,
                          title_text="Multiple Histogram for all featrues")



def multiple_boxplot(data):
    
    num_row = math.ceil(len(data.columns) / 3)
    fig = make_subplots(rows=num_row, cols=3,subplot_titles=(df.columns))

    place_col = 1
    place_row = 1
    for i in data.columns:

        fig.add_trace(go.Box(y=data[i]),row=place_row, col=place_col)

        place_col += 1
        if place_col == 4:
            place_col = 1
            place_row += 1
    return fig.update_layout(height=1600, width=1100,
                          title_text="Multiple Histogram for all featrues")

def check_nulls(X_data):
    
    unusual_nulls = X_data[X_data.isin(['{}','[]', "?", ".", "-", "_", "", " ", "  "])].sum()

    nulls_df = pd.concat([X_data.isna().sum(), unusual_nulls], axis=1)
    nulls_df.columns = ["usual_nulls", "strange_nulls"]
    nulls_df = nulls_df.sort_values('usual_nulls', ascending = False)
    return nulls_df

########################################################

def kMeansInitCentroids(X, K):
   
    centroids=np.zeros((K, X.shape[1]))
    sample_size = int(X.shape[0]/K)
    
    for i in range(K):
        
        index = np.random.randint(X.shape[0], size=sample_size)
        
        new_centriod = sum(X[index])/sample_size
        centroids[i, :]=new_centriod
   
    return centroids


def findClosestCentroids(X, centroids):
    
    matrix=np.zeros((X.shape[0], centroids.shape[0]))
    for i in range(centroids.shape[0]):
        
        diffrences = X-centroids[i]
        norm_vector = np.linalg.norm(diffrences, axis=1)
        
        matrix[:,i]=norm_vector
    idx=np.argmin(matrix, axis=1).reshape(-1,1)
    
    return idx


def computeCentroids(X, idx, K, centroids_old):
    
    centroids=np.zeros((K, X.shape[1]))
    idx = idx.reshape(X.shape[0])
    
    for i in range(K):
        if len(X[idx==i]) ==0:
            centroids[i] = centroids_old[i]
        else:
            centroids[i]=np.mean(X[idx==i], axis=0)

    return centroids

def k_means(X,K):
    
    X_data = pd.DataFrame(X)
    
    list_of_idx=[]
    list_of_centroids=[]
    list_of_sse=[]
    
    for j in range(100):
        centroids = kMeansInitCentroids(X, K)
        inertia = 0

        for i in range(300):

            idx = findClosestCentroids(X, centroids)
            centroids = computeCentroids(X, idx, K, centroids)

        for i in range(K):
            if len(X_data[idx==i]) == 0:
                continue
            else:
                diff= X_data[idx==i]-centroids[i]
                inertia += sum(np.square(norm(diff, axis=1)))
                
        list_of_idx.append(idx)
        list_of_centroids.append(centroids)
        list_of_sse.append(inertia)
    
    min_sse_index = np.argmin(list_of_sse)
    
    return list_of_centroids[min_sse_index], list_of_idx[min_sse_index], min(list_of_sse)



def elbow(X, K):
    
    import plotly.express as px
    variances=[]
    for i in range(1, K+1):
        
        centroids, idx, variance = k_means(X, i)
        variances.append(variance)
        
    return px.line(x=np.arange(1,K+1), y=variances, markers=True,labels ={'x':"number of cetriods", 'y':"Inertia"},
                                       range_y=[0, max(variances)+200] ,title="Sum of variances VS number of Cenriods in agiven data set")

def predict(centroids,X):
    
    matrix=np.zeros((X.shape[0], centroids.shape[0]))
    
    for i in range(centroids.shape[0]):
        
        diffrences = X-centroids[i]
        norm_vector = np.linalg.norm(diffrences, axis=1)
        
        matrix[:,i]=norm_vector
    label=np.argmin(matrix, axis=1).reshape(-1,1)

    return label

Data Loading & EDA

#read the data
df = pd.read_csv('CC GENERAL.csv')
df.drop('CUST_ID', axis=1, inplace=True)
print('The shape of the dataset is:', df.shape)

The shape of the dataset is: (8950, 17)

df.head()

	BALANCE	BALANCE_FREQUENCY	PURCHASES	ONEOFF_PURCHASES	INSTALLMENTS_PURCHASES	CASH_ADVANCE	PURCHASES_FREQUENCY	ONEOFF_PURCHASES_FREQUENCY	PURCHASES_INSTALLMENTS_FREQUENCY	CASH_ADVANCE_FREQUENCY	CASH_ADVANCE_TRX	PURCHASES_TRX	CREDIT_LIMIT	PAYMENTS	MINIMUM_PAYMENTS	PRC_FULL_PAYMENT	TENURE
0	40.900749	0.818182	95.40	0.00	95.4	0.000000	0.166667	0.000000	0.083333	0.000000	0	2	1000.0	201.802084	139.509787	0.000000	12
1	3202.467416	0.909091	0.00	0.00	0.0	6442.945483	0.000000	0.000000	0.000000	0.250000	4	0	7000.0	4103.032597	1072.340217	0.222222	12
2	2495.148862	1.000000	773.17	773.17	0.0	0.000000	1.000000	1.000000	0.000000	0.000000	0	12	7500.0	622.066742	627.284787	0.000000	12
3	1666.670542	0.636364	1499.00	1499.00	0.0	205.788017	0.083333	0.083333	0.000000	0.083333	1	1	7500.0	0.000000	NaN	0.000000	12
4	817.714335	1.000000	16.00	16.00	0.0	0.000000	0.083333	0.083333	0.000000	0.000000	0	1	1200.0	678.334763	244.791237	0.000000	12

# Let's see the data types and non-null values for each column
df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 8950 entries, 0 to 8949
Data columns (total 17 columns):
 #   Column                            Non-Null Count  Dtype  
---  ------                            --------------  -----  
 0   BALANCE                           8950 non-null   float64
 1   BALANCE_FREQUENCY                 8950 non-null   float64
 2   PURCHASES                         8950 non-null   float64
 3   ONEOFF_PURCHASES                  8950 non-null   float64
 4   INSTALLMENTS_PURCHASES            8950 non-null   float64
 5   CASH_ADVANCE                      8950 non-null   float64
 6   PURCHASES_FREQUENCY               8950 non-null   float64
 7   ONEOFF_PURCHASES_FREQUENCY        8950 non-null   float64
 8   PURCHASES_INSTALLMENTS_FREQUENCY  8950 non-null   float64
 9   CASH_ADVANCE_FREQUENCY            8950 non-null   float64
 10  CASH_ADVANCE_TRX                  8950 non-null   int64  
 11  PURCHASES_TRX                     8950 non-null   int64  
 12  CREDIT_LIMIT                      8949 non-null   float64
 13  PAYMENTS                          8950 non-null   float64
 14  MINIMUM_PAYMENTS                  8637 non-null   float64
 15  PRC_FULL_PAYMENT                  8950 non-null   float64
 16  TENURE                            8950 non-null   int64  
dtypes: float64(14), int64(3)
memory usage: 1.2 MB

check_nulls(df)

	usual_nulls	strange_nulls
MINIMUM_PAYMENTS	313	0.0
CREDIT_LIMIT	1	0.0
BALANCE	0	0.0
CASH_ADVANCE_FREQUENCY	0	0.0
PRC_FULL_PAYMENT	0	0.0
PAYMENTS	0	0.0
PURCHASES_TRX	0	0.0
CASH_ADVANCE_TRX	0	0.0
PURCHASES_INSTALLMENTS_FREQUENCY	0	0.0
BALANCE_FREQUENCY	0	0.0
ONEOFF_PURCHASES_FREQUENCY	0	0.0
PURCHASES_FREQUENCY	0	0.0
CASH_ADVANCE	0	0.0
INSTALLMENTS_PURCHASES	0	0.0
ONEOFF_PURCHASES	0	0.0
PURCHASES	0	0.0
TENURE	0	0.0

# This will print basic statistics for numerical columns
df.describe()

	BALANCE	BALANCE_FREQUENCY	PURCHASES	ONEOFF_PURCHASES	INSTALLMENTS_PURCHASES	CASH_ADVANCE	PURCHASES_FREQUENCY	ONEOFF_PURCHASES_FREQUENCY	PURCHASES_INSTALLMENTS_FREQUENCY	CASH_ADVANCE_FREQUENCY	CASH_ADVANCE_TRX	PURCHASES_TRX	CREDIT_LIMIT	PAYMENTS	MINIMUM_PAYMENTS	PRC_FULL_PAYMENT	TENURE
count	8950.000000	8950.000000	8950.000000	8950.000000	8950.000000	8950.000000	8950.000000	8950.000000	8950.000000	8950.000000	8950.000000	8950.000000	8949.000000	8950.000000	8637.000000	8950.000000	8950.000000
mean	1564.474828	0.877271	1003.204834	592.437371	411.067645	978.871112	0.490351	0.202458	0.364437	0.135144	3.248827	14.709832	4494.449450	1733.143852	864.206542	0.153715	11.517318
std	2081.531879	0.236904	2136.634782	1659.887917	904.338115	2097.163877	0.401371	0.298336	0.397448	0.200121	6.824647	24.857649	3638.815725	2895.063757	2372.446607	0.292499	1.338331
min	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	50.000000	0.000000	0.019163	0.000000	6.000000
25%	128.281915	0.888889	39.635000	0.000000	0.000000	0.000000	0.083333	0.000000	0.000000	0.000000	0.000000	1.000000	1600.000000	383.276166	169.123707	0.000000	12.000000
50%	873.385231	1.000000	361.280000	38.000000	89.000000	0.000000	0.500000	0.083333	0.166667	0.000000	0.000000	7.000000	3000.000000	856.901546	312.343947	0.000000	12.000000
75%	2054.140036	1.000000	1110.130000	577.405000	468.637500	1113.821139	0.916667	0.300000	0.750000	0.222222	4.000000	17.000000	6500.000000	1901.134317	825.485459	0.142857	12.000000
max	19043.138560	1.000000	49039.570000	40761.250000	22500.000000	47137.211760	1.000000	1.000000	1.000000	1.500000	123.000000	358.000000	30000.000000	50721.483360	76406.207520	1.000000	12.000000

multiple_histogram(df)

multiple_boxplot(df)

plt.figure(figsize = (14,8))
sns.heatmap(df.corr(),annot=True,linewidths=0.1,  cmap= 'RdBu');

#make a copy for the original dataset
df_copy=df.copy()

#solution 
df_copy.dropna(axis =0, inplace =True )
df_copy.isnull().sum()

BALANCE                             0
BALANCE_FREQUENCY                   0
PURCHASES                           0
ONEOFF_PURCHASES                    0
INSTALLMENTS_PURCHASES              0
CASH_ADVANCE                        0
PURCHASES_FREQUENCY                 0
ONEOFF_PURCHASES_FREQUENCY          0
PURCHASES_INSTALLMENTS_FREQUENCY    0
CASH_ADVANCE_FREQUENCY              0
CASH_ADVANCE_TRX                    0
PURCHASES_TRX                       0
CREDIT_LIMIT                        0
PAYMENTS                            0
MINIMUM_PAYMENTS                    0
PRC_FULL_PAYMENT                    0
TENURE                              0
dtype: int64

df_logged = np.log1p(df_copy)

from sklearn.manifold import TSNE

tsne_projection = TSNE(n_components=2, 
                       perplexity=30, 
                       learning_rate=1, 
                       init='pca', 
                       method='barnes_hut', 
                       n_jobs=-1, 
                       n_iter=10**4, 
                       random_state=42).fit_transform(df_logged)

from sklearn.decomposition import PCA

pca_logged = PCA()
df_logged_red = pca_logged.fit_transform(df_logged)

d = {'Number of EigenVector': np.arange(1,len(df_copy.columns)+1), 'Varaiability Captured': np.cumsum(pca_logged.explained_variance_ratio_)}
df = pd.DataFrame(data=d)
px.line(df, x='Number of EigenVector', y='Varaiability Captured',markers=True,
                       title='EigenVectors VS Varaiability Captured')

pca_logged = PCA(n_components=8)
df_logged_red = pca_logged.fit_transform(df_logged)

from sklearn.neighbors import NearestNeighbors
from sklearn.cluster import DBSCAN

# we ask for mins + 1 nearest, because the data point it self (distance = 0) is included
mins = 5
knn = NearestNeighbors(mins + 1)

knn.fit(df_logged_red)

distances, neighbors = knn.kneighbors(df_logged_red)

# sort the distances
distances = np.sort(distances[:, mins], axis = 0)
distances

array([0.19741106, 0.20389162, 0.2173771 , ..., 4.3076648 , 4.53674092,
       6.14210845])

#plot the distances
distances_df = pd.DataFrame({"distances": distances,
                             "index": list(range(0,len(distances)))})
plt = (ggplot(distances_df, aes(x = "index", y = "distances")) +
 geom_line(color = "white", size = 2) + theme_minimal() +
 labs(title = "Elbow Method for Choosing eps") +
 theme(panel_grid_minor = element_blank(),
      rect = element_rect(fill = "#202124ff"),
      axis_text = element_text(color = "white"),
      axis_title = element_text(color = "white"),
      plot_title = element_text(color = "white"),
      panel_border = element_line(color = "darkgray"),
      plot_background = element_rect(fill = "#202124ff")
      ))
ggsave(plot=plt, filename='elbow.png', dpi=300)

plt

<ggplot: (194910081034)>

db1 = DBSCAN(eps = 1.8, min_samples = 5).fit(df_logged_red)
labels = db1.labels_
silhouette_score(df_logged_red, labels)

0.46007490327294787

db1 = DBSCAN(eps = 1.8, min_samples = 5).fit(df_logged_red)
tsne_projection = pd.DataFrame(tsne_projection, columns=['feature1', 'feature2'],index=df_copy.index.tolist())
tsne_projection['cluster'] = db1.labels_
tsne_projection.replace(-1, "anamoly", inplace=True)
px.scatter(tsne_projection, x='feature1', y='feature2', color="cluster")

Kernal PCA

from sklearn.decomposition import KernelPCA

kpca = KernelPCA(n_components = 8,kernel='rbf',gamma=0.02)
df_logged_red = kpca.fit_transform(df_logged)

# we ask for mins + 1 nearest, because the data point it self (distance = 0) is included
mins = 5
knn = NearestNeighbors(mins + 1)

knn.fit(df_logged_red)

distances, neighbors = knn.kneighbors(df_logged_red)

# sort the distances
distances = np.sort(distances[:, mins], axis = 0)
distances

array([0.00794466, 0.00840996, 0.00869148, ..., 0.34386759, 0.34506282,
       0.35687502])

#plot the distances
distances_df = pd.DataFrame({"distances": distances,
                             "index": list(range(0,len(distances)))})
plt = (ggplot(distances_df, aes(x = "index", y = "distances")) +
 geom_line(color = "white", size = 2) + theme_minimal() +
 labs(title = "Elbow Method for Choosing eps") +
 theme(panel_grid_minor = element_blank(),
      rect = element_rect(fill = "#202124ff"),
      axis_text = element_text(color = "white"),
      axis_title = element_text(color = "white"),
      plot_title = element_text(color = "white"),
      panel_border = element_line(color = "darkgray"),
      plot_background = element_rect(fill = "#202124ff")
      ))
ggsave(plot=plt, filename='elbow.png', dpi=300)

plt

<ggplot: (194920885411)>

db1 = DBSCAN(eps = 0.15, min_samples = 5).fit(df_logged_red)
labels = db1.labels_
silhouette_score(df_logged_red, labels)

0.5418912605026349

db1 = DBSCAN(eps = 0.15, min_samples = 5).fit(df_logged_red)
tsne_projection = pd.DataFrame(tsne_projection, columns=['feature1', 'feature2'],index=df_copy.index.tolist())
tsne_projection['cluster'] = db1.labels_
tsne_projection.replace(-1, "anamoly", inplace=True)
px.scatter(tsne_projection, x='feature1', y='feature2', color="cluster")